The University of Hong Kong
BSc · Mathematics & Computer Science
Mathematics × Computer Science
Stanley Chow.
Mathematics & Computer Science student at the University of Hong Kong
I work on scientific computing and machine learning, with current projects in inverse problems and sharpness-aware optimization.
Career interests: Data science · ML engineering · Recommender systems · Quant
Research in motion · HKU Summer Research Fellowship
Finding a hidden heat source.
Heat spreads and blurs a source’s shape. Can we work backwards from the temperature to find where it came from?
Watch three models tackle the same problem. A smaller model can finish quickly, but the examples it learns from affect the shape it recovers.
Watch the 34-second demo
34 seconds · No audio Research & current work
Research overviewCapstone: sharpness-aware optimization
In progressI’m exploring sharpness-aware optimization for machine-learning training. The work is ongoing; there are no published results yet.
HKU Summer Research Fellowship
2026I study how to reconstruct a hidden heat source, how to judge its recovered shape, and when a reduced model can make the computation faster. I was selected as one of the best poster presenters and invited to present at the HKU Science Undergraduate Research Colloquium (Summer 2026).
Explore the researchUrban rail network optimization
2023I modeled rail routes under geometric and engineering constraints, combining analytical optimization with Particle Swarm Optimization.
Education
University of California, Los Angeles
Exchange study · Mathematics & Computer Science
Stanford University
Visiting study · Mathematics & Computer Science
Selected projects
All projectsRecommendation systems
Personalized recommendations for H&M
A two-stage recommender built from millions of shopping transactions. It combines nine retrieval channels, then uses a learned model to rank the candidates.
31.8 million transaction events
Search & software engineering
DocuQuest Agent
A document question-answering system with a C++ search layer. I built its index and retrieval logic, with deterministic tests that run without API credentials.
Custom index · Reproducible tests
Machine learning
Behavioral personality analytics
A study of how behavioral features predict personality outcomes, with careful separation of model selection and final evaluation.
381 configurations · Generated-data study
Blog
All postsProject write-up
HM Recommendations: Find! Then, Rank!
How I built a two-stage recommender from 31.8 million purchases, tested it on a later week, and produced recommendations for 1.37 million customers.
Study notes
Learning recommendation systems with Wang Shusen
A short guide to Wang Shusen's lectures, with my Chinese study notes to read alongside the course.
Research notes
Finding a shape after heat has spread
A short introduction to my inverse-problem research: working backwards from temperature measurements to find a hidden heat source.
Tools & methods
Recommendation & search
Candidate retrieval, collaborative filtering, two-tower models, LightGCN, LambdaRank, negative sampling, and temporal validation.
Machine learning & data
Python, SQL, PyTorch, LightGBM, XGBoost, scikit-learn, pandas, NumPy, SciPy, and DuckDB.
Mathematics & engineering
C++, MATLAB, numerical optimization, partial differential equations, experiment design, Git, CMake, Jupyter, and LaTeX.
Let’s connect
Have an idea worth exploring?
I’m looking for opportunities in data science, machine learning engineering, recommendation systems, and quant. I’m also happy to discuss scientific computing and inverse problems.
pingsenchow@outlook.com